Message/Author 

Anonymous posted on Monday, October 03, 2005  3:38 pm



This is my first foray into the WRMSR. What would be a standard for a good WRMSR? My model produced one that is 3.19. Intuiatively, this is very high to me. How could I judge and are there any citations that I can refer too? 

bmuthen posted on Saturday, October 08, 2005  2:02 pm



See the Yu dissertation that is available on the Mplus web site. A value less than 1 is generally good, but note that WMRMR has been found to not work well for growth models and multiplegroup models. 


How does WMRMR work with MIMIC models where the exogenous varaibels are either metric or dummy coded? Overall my model seems to congerge ok (Note: N>4K) Value 485.324* Degrees of Freedom 39 PValue 0.0000 Scaling Correction Factor 1.256 for MLM CFI 0.975 TLI 0.967 RMSEA 0.053 SRMR 0.024 WMRMR 1.924 The only problem seems to be WMRMR 


I would not worry about WMRMR. It is not a wellstudied fit statistic and has not behaved as well as expected. 


Hello, I have run a path model. I had these fit statistics values: RMSEA: 0.02 CFI: 0.999 WRMR (Weighted Root Mean Square Residual): 0.3 My question is what is the difference between (Weighted Root Mean Square Residual) and (Standardized Root Mean Square Residual (SRMR))? and how can I get SRMR in my output? Thank you, Owis 


SRMR is not available for WLSMV if the model has thresholds or covariates. The default is to include thresholds in the model. If you do not have covariates, you can use MODEL = NOMEANSTRUCTURE; in the ANALYSIS command. 


Thank you for the reply. I already used MODEL = NOMEANSTRUCTURE in my input. If I dont misunderstand you, SRMR is not available for WLSMV if the model has categorical outcomes! Thank you, Owis 


If you use MODEL = NOMEANSTRUCTURE and have no covariates, you should get SRMR. If you don't, send the output and your license number to support@statmodel.com. 

Back to top 